This project aims at evaluating the treatment selection and prognostic information contained in the sleep polysomnograph (PSG), dexamethasone suppression test (DST), and TRh stimulation test (TRHST) in unipolar, nonpsychotic major depressions. 100 subjects, 50 on trazodone and 50 on desipramine, will be studied prior to and during a randomized double-blind trial lasting 6 weeks (acute phase), during the drug maintenance phase (6 months), and following taper to a drug-free status for the subsequent 12 months. By design, at least 1/2 of each drug treatment group will have a reduced (less than 60 minutes) REM latency. The patients will also complete psychologic measures of attitudes, cognitions, and attributions at each assessment point. The study will determine if (a) pretreatment PSG, DST, or TRHST status predicts overall drug responsiveness or differential drug response; (b) pretreatment or past-acute phase laboratory test status predicts symptomatology during the maintenance phase; (c) pretreatment, post-acute phase, or post-drug taper laboratory tests predict relapse during the drug-free followup and (d) whether these laboratory tests add to or exceed predictions of acute drug response, maintenance phase symptomalology, or drug-free phase relapse derived from clinical variables or psychologic test performance. Further the relative state vs trait qualities of each laboratory and psychologic test will be discerned. Finally, as repeated monthly laboratory and clinical assessments during the drug-free followup will be conducted, a picture of the relationship between these tests and the development of clinical relapse will be clarified. That is, a model of which tests parameters are pathogenetically more proximal to the underlying derangements in depression can be developed.